
Another update, and this one didn't take nearly a year! Sensational! This new version is trained on Pony Diffusion v6, using pretty much the same dataset as V2. PDXL already knows what Penny looks like, but not her clothes. That probably makes about half the dataset redundant, but whatever, it works.
描述:
10 months later, I've finally updated this thing. And hoo boy did it need updating! V1 was my first LORA, and it showed. V2's tagging is more consistent, it is trained on NAI instead of AnythingV3, plays much nicer with style LORAs, and doesn't melt her eyeballs anymore. The dataset is also 4x larger and features more varied angles, NSFW, and alternate outfits. You no longer have to prompt for orange hair, ahoge, or green eyes (usually), and can add more weight to "curly hair" if the model you're using doesn't understand her cinnamon bun curls.
V2 is trained at 1024x1024, which is high for an SD1.5 model and can sometimes cause issues with highres fix, but seems to preserve details better. I've also uploaded a 768x768 version for those who prefer it.
Some of the preview images aren't one-shot txt2img, but the metadata should give you the base I worked from.
训练词语: Penny Polendina, curly hair, freckles,,short hair, pink bow, gorget, black thighhighs, neon trim, gold trim, short dress, grey dress, underbust, pencil skirt, white blouse,,long hair, black bow, neck ribbon, corset, suspender skirt, white blouse, mechanical legs,,Floating Array, sword, backpack,,doll joints,
名称: PennyPolendinaV2-1024.safetensors
大小 (KB): 36991
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success